I’ve integrated tempest with our loxone home automation system easily using an existing loxberry plugin. Works great.
However, to predict when to start cooling the home, it would be nice to have a prediction of the expected solar radiation. The uv-prediction is already there and has a correlation seemingly, but it would be nice if you could add this in a future release, as the current solar radiation is recorded by the weather system.
I’m not completely sure, but a prediction of 3 days in advance would be nice. It depends on the accuracy of course, and the amount of overheating/heating. Modern houses don’t cool/heat as fast due to strong thermal isolation and low-temperature heating, so a longer forecast seems useful.
I’ll check the loxone-setup again as it already expects a certain amount of data, but it requires a per hour forecast to manage screens, …
I already added the night time solar radiation :).
Thank you and kind regards,
Where are getting UV prediction from? I don’t see that in my forecast at all.
As for solar radiation and heating and cooling your house is complicated.
I have been working on that myself but only for the current day so I can determine if it is cloudy.
I’m curious about your thought process for your heating and cooling based on solar radiation (SR). Are you assuming a certain external heating rate of your house based on a given SR?
I am up to 5 different SR estimates that are more accurate at different solar angles. I, unfortunately, do not have solid data as I have had clouds for weeks and I need a few days of pure sun to test my assumptions.
Well the loxone automation system does that for me. It just expects the expected external solar radiation where I live. I do not think it is very complicated as the processing power is limited. However, I suspect that based on internal temperature sensors and state of screens (protection) one can estimate pretty well the effect of the solar radiation. No need to know every detail of your home I think, if it can be guessed with 90% accuracy.
Sorry, forgot. The Uv prediction is present in the hourly prediction of the webapi.
never knew it there was UV prediction. Does this take into account the clouds? is it the average or the maximum of the hour? What would you need for solar radiation also the maximum expected? Or do you want to guess the amount of energy your solar panels produce (that would need to include a prediction of cloud cover, as it is integrated over time).
Well, based on my experience, the uv index is based on what arrives on your skin, so taking clouds into account. On a cloudy day, it seems less, although I’m just started to look at these values recently.
solar radiation (W/m²) should be the average over an hour I suppose, so that it actually represents Wh/m². Seems the most logical thing to be honest. The goal is to express how much energy from the sun is arriving each hour at that specific spot.
I’m interested in seeing the reduction of UV/Brightness/Solar Radiation from their maximum theoretical values. It would be ideal if a theoretical curve could be overlaid on each of these graphs so the reduction would be easily seen, but even a single number would be helpful. I would expect that the theoretical curve would be dependent on the Latitude and the time of year.
Any comments or suggestions?
Something like this?
I decided to test out multiplying uv index times max solar radiation. Should get me close, before starting on a neural net
Working on it:
I am actually up to 6 different equations I am testing out. I have not had a non-cloudy day in months so I can not fully test myself until that time. I think I have it pretty good in that when there is not a cloud overhead at least one of my equations is fairly close but I can’t be for certain. I also need to trim up the area where it is effective because around 80 degrees Solar Azimuth (20 degrees Solar Elevations) the light sensor on the Tempest does not pick up to much. I can post my functions if anyone would like to test for themselves. Some of them take into account station height and Air Mass of the distance through air between the station and the sunlight entering the atmosphere.
I have the outputs feed back to Home Assistant so I can graph them easier. So I can post better screenshots also. I also run a comparison of ‘expected’ Solar Radiation to measured Solar Radiation from the Tempest; this is what I used currently to determine if actually cloudy. If the expected is twice the measured then it is cloudy; this seems to be working well…but as stated I need a full sunny day to be able to dial in my functions.
My six estimates on Solar Radiation:
Left Column______________ Right Column
Solar Elevation____________Solar Zenith
Measured SR_____________Measured SR
Estimate 1_______________ Comparison of Est to Measured in Percent
Estimate 2_______________ Comparison of Est to Measured in Percent
Estimate 3_______________ Comparison of Est to Measured in Percent
Estimate 4_______________ Comparison of Est to Measured in Percent
Estimate 5_______________ Comparison of Est to Measured in Percent
Estimate 6_______________ Comparison of Est to Measured in Percent
I’m curious as to what you come up with. As you can see I’m testing six different ones; my 4th seems to be the most accurate for now. I can post my equations if you would like to see what I am testing. It would be nice to compare to other geographic locations too.